Graphics Reference
In-Depth Information
smoothing spline method (Gu,
)is currently under investigation as a method to
obtain the mode surface.
Genetic networks of
genes in yeast that are SSL to SGS
orRAD
were of
interest. SGS
(RAD
) has homologs in human cells, including WRN, BLM and
RECQ
(FEN
andERCC
) genes. Mutations in these genes lead to cancer-pre-
disposition syndromes, symptoms resembling premature aging, and Cockayne syn-
drome (Tong et al.,
and NCBI OMIM database). he regression approach was
applied to these
genes to infer their TC interactions using the alpha data set in
Spellman et al. (
). here were
quartets ( β
, β
, β
, β
) that satisfied the cri-
teria
values of β
, β
, β
and β
R
.
and p
−
<
.
.
(
.
)
Uponpartitioning the
quartets into
cubesaccordingtoSilverman's
rule, Shieh and her colleagues obtained the following surface:
T
i
(
t
+
)=
.
+
.
A
i
(
t
)−
.
R
i
(
t
)+
.
A
i
(
t
)
R
i
(
t
)
(
.
)
Fromthetoptentriplets, asmallnetworkwasreconstructed.SomeTCandTDinter-
actions were consistent with the published literature, so the prediction was promis-
ing. Toinfernovel TC interactions, the Scorewasrelaxed to
.
and
triplets were
included,whichresultedinalargernetwork.Amongthosepredictions,someTCand
TD interactions of interest were checked by RT-PCR experiments; the prediction ac-
curacy including both layers was about
%. Note that the mode surface is deter-
mined by the majority of triplets, so this regression approach can be applied to any
data set. Figure
.
confirms not only the validity of taking a data-driven approach,
but also the development of the mode surface.
A Pattern Recognition Approach
1.3.5
SimilarlytoSect.
.
,transcriptional compensation (TC)interactions amongagroup
of
yeast genes which are synthetic sick or lethal to SGS
orRAD
(Tong et al.,
) are also of interest. Chuang et al. (
) proposed a pattern recognition ap-
proach to infer TC interactions. he proposed approach, in fact, was implemented
on indirect interactions among RT-PCR-confirmed TC and transcriptional dimin-
ishment (TD) gene pairs. For ease of description, in this section we utilize A, R and
T, which are involved in direct interactions, to denote the genes involved in TD and
TC interactions. Among RT-PCR-confirmed gene pairs, when the time course mi-
croarray gene expression (Spellman et al.,
)of a target gene T was plotted lagged
−
in time behind that of A or R, the AT gene pairs generally exhibited similar pat-
terns across time, as depicted in the blue zone of Fig.
.
. On the other hand, RT
gene pairs showed complementary patterns across time, as illustrated in Fig.
.
.
his is consistent with the ideal patterns for RT and AT, in which the expression of R
and that of its corresponding lagged
−
T are mirror images, whereas those of A and
lagged
T are similar. Furthermore, the areas enclosed by the RT curves are much
larger than those enclosed by the AT curves, and this phenomenon was displayed
−
